1091generated behavioral tests
79.5%official best score
79.7%Codex best score
3Codex result rows
Codex Results by Model
| # | Model | Run | Agent | Mode | Score | Eval | Tests | Est. cost | Calls | Wall | Evidence |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | 20260517T094700Z |
Codex /goal | Mini-SWE-compatible no internet | 79.7% | ok | 870/1091 | $2.76 | 43 | 0.15h | manifest · eval json · eval summary · usage audit |
| 2 | GPT 5.5 (xhigh) | 20260519T215046Z |
Codex /goal | Paper prompt + Goal contract no internet | 75.2% | ok | 820/1091 | n/a | 36 | 0.16h | not exported |
| 3 | GPT 5.5 (xhigh) | 20260518T232006Z |
Codex /goal | Paper prompt + /goal no internet | 73.9% | ok | 806/1091 | n/a | 44 | 0.12h | not exported |
Why Scores Differ
Official ProgramBench rows are the public mini-SWE-agent submissions. GoalBench rows are separate Codex /goal submissions. Same model label does not mean the same agent, prompt, tool loop, or generated implementation. A GoalBench row is resolved only when the ProgramBench-scored pass rate is exactly 100%.
GPT 5.5 (xhigh) · 20260517T094700Z
79.7% from 870/1091 ProgramBench-scored tests. Raw public eval statuses: failure: 310, passed: 1139, skipped: 13.
Agent trace summary unavailable.
Likely miss class: behavioral mismatch, exact version/output mismatch.
b085a6cde326eval.tests.test_additional_coverage.test_multiple_values_from_files (failure)b085a6cde326eval.tests.test_additional_coverage.test_value_from_file (failure)b085a6cde326eval.tests.test_bench_comprehensive.test_bench_with_method_override (failure)b085a6cde326eval.tests.test_download_mode.test_download_basic (failure)b085a6cde326eval.tests.test_download_mode.test_download_large_file (failure)b085a6cde326eval.tests.test_download_mode.test_download_with_auth (failure)
GPT 5.5 (xhigh) · 20260519T215046Z
Public eval evidence has not been exported for this row yet.
GPT 5.5 (xhigh) · 20260518T232006Z
Public eval evidence has not been exported for this row yet.
Official ProgramBench Results by Model
| # | Model | Provider | Score | Cost | Calls |
|---|---|---|---|---|---|
| 1 | GPT 5.5 (xhigh) | OpenAI | 79.5% | $5.83 | 49 |
| 2 | GPT 5.5 | OpenAI | 79.4% | $1.16 | 17 |
| 3 | GPT 5.5 (high) | OpenAI | 78.2% | $2.19 | 21 |
| 4 | Claude Opus 4.6 | Anthropic | 71.8% | $9.63 | 188 |
| 5 | Claude Opus 4.7 (xhigh) | Anthropic | 70.9% | $13.25 | 132 |
| 6 | Claude Opus 4.7 | Anthropic | 66.4% | $4.95 | 126 |
| 7 | Gemini 3 Flash | 65.4% | $0.29 | 85 | |
| 8 | Claude Sonnet 4.6 | Anthropic | 64.5% | $18.12 | 365 |
| 9 | GPT 5.4 | OpenAI | 58.9% | $0.17 | 9 |
| 10 | Gemini 3.1 Pro | 50.5% | $1.25 | 55 | |
| 11 | Claude Haiku 4.5 | Anthropic | 49.9% | $0.58 | 88 |
| 12 | GPT 5.4 mini | OpenAI | 26.4% | $0.04 | 8 |
| 13 | GPT 5 mini | OpenAI | 16.0% | $0.01 | 9 |